Subunit NDUFV3 Is Present in Two Distinct Isoforms in Mammalian Complex I

Biochimica et Biophysica Acta 1858 (2017) 197–207

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

journal homepage: www.elsevier.com/locate/bbabio

Subunit NDUFV3 is present in two distinct isoforms in mammalian complex I

Hannah R. Bridges 1, Khairunnisa Mohammed 1, Michael E. Harbour, Judy Hirst ⁎

The Medical Research Council Mitochondrial Biology Unit, Wellcome Trust / MRC Building, Hills Road, Cambridge, CB2 0XY, U. K. article info abstract

Article history: Complex I (NADH:ubiquinone oxidoreductase) is the first enzyme of the electron transport chain in mammalian Received 27 October 2016 mitochondria. Extensive proteomic and structural analyses of complex I from Bos taurus heart mitochondria have Received in revised form 29 November 2016 shown it comprises 45 subunits encoded on both the nuclear and mitochondrial genomes; 44 of them are differ- Accepted 7 December 2016 ent and one is present in two copies. The bovine heart enzyme has provided a model for studying the composition Available online 08 December 2016 of complex I in other mammalian species, including humans, but the possibility of additional subunits or isoforms fi Keywords: in other species or tissues has not been explored. Here, we describe characterization of the complexes I puri ed fi Complex I from ve rat tissues and from a rat hepatoma cell line. We identify a ~ 50 kDa isoform of subunit NDUFV3, for isoform which the canonical isoform is only ~10 kDa in size. We combine LC-MS and MALDI-TOF mass spectrometry mitochondria data from two different purification methods (chromatography and immuno-purification) with information NADH:ubiquinone oxidoreductase from blue native PAGE analyses to show the long isoform is present in the mature complex, but at NDUFV3 substoichiometric levels. It is also present in complex I in cultured human cells. We describe evidence that the rat long isoform is more abundant in both the mitochondria and purified complexes from brain (relative to in heart, liver, kidney and skeletal muscle) and more abundant still in complex I in cultured cells. We propose that the long 50 kDa isoform competes with its canonical 10 kDa counterpart for a common binding site on the flavoprotein domain of complex I. © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).

1. Introduction have vastly increased the availability of genetic data in recent years [3], but interpretation of the links between specific mutations and clin- Respiratory complex I (NADH:ubiquinone oxidoreductase) is the ical or pathological phenotypes still relies heavily on basic knowledge of first enzyme of the electron transport chain in mammalian mitochon- the identities and sequences of the proteins involved. dria [1]. It oxidizes NADH in the mitochondrial matrix to regenerate Complex I isolated from Bos taurus (bovine) heart mitochondria is NAD+ and sustain crucial metabolic processes including the tricarboxyl- the most comprehensively studied mammalian complex I, and its ic acid cycle and β-oxidation of fatty acids, reduces ubiquinone in the subunit composition has been used as the model for the human enzyme. inner membrane to supply electrons to respiratory complex III, and During the 1980s and 1990s Walker and coworkers identified 43 transports protons across the membrane, contributing to the proton proteins in preparations of the bovine enzyme and its subcomplexes, motive force that drives ATP synthesis and transport processes. Com- and sequenced 35 different nuclear-encoded proteins [4–6].Seven plex I is also a significant source of mitochondrial reactive oxygen spe- additional subunits were found to be encoded in the mitochondrial cies production and so contributes to cellular oxidative stress. Due to genome, making a total of 42 different sequences at this time [7].There- its critical contribution to cellular metabolism, dysfunctions of complex maining protein was subsequently found to be an unusual fragment of I are the most frequent causes of mitochondrial disease [2].ComplexI one of the known subunits [8]. Then, in an extensive re-evaluation of defects are clinically and phenotypically diverse, and diagnosis of genet- the enzyme's subunit composition using state-of-the-art mass spec- ically-linked complex I dysfunctions, caused by mutations in both the trometry methods, three more subunits were identified, giving a total mitochondrial and nuclear subunits of the enzyme, and in the assembly of 45 different sequences [9,10]. One of these proteins (NDUFA4), al- factors required for its biogenesis, relies on both biochemical and genet- ways in doubt as a bona fide subunit due to its weak association with ic information. Advances in high-throughput sequencing techniques complex I and its presence in more than one chromatographic fraction [11], has now been discounted as a complex I subunit [12].Theremain- ing 44 different subunits have been confirmed by determination of the ⁎ Corresponding author. E-mail address: [email protected] (J. Hirst). structure of the bovine enzyme, and this also revealed that one subunit, 1 These authors contributed equally. the mitochondrial acyl-carrier protein, is present in two copies [13,14].

http://dx.doi.org/10.1016/j.bbabio.2016.12.001 0005-2728/© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). 198 H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207

Therefore, bovine heart complex I is currently understood to contain 45 glycerol and 1 mM EDTA. Membranes were prepared using a protocol subunits in total. Fourteen of the subunits of bovine complex I are the adapted from that of Walker et al. [23] by addition of solid KCl to catalytic ‘core’ subunits that are conserved in all species of complex I 0.15 M followed by electrical homogenization, and collected by centri- and contain the mechanistic elements sufficient to catalyze NADH oxi- fugation (13,500 x g for 40 min). dation, ubiquinone reduction and proton translocation [1]. The additional 31 subunits present are ‘supernumerary’ or ‘accessory’ subunits 2.3. Isolation of mitochondria from U2OS cells [11]. They have been accumulated onto the core during evolution, and both their number and nature vary widely between species [15]. Mitochondria were isolated from cultured cells using a protocol In general, the composition of complex I from other mammalian spe- adapted from that of Minczuk et al. [24]. U2OS cells (~1.5 g wet weight) cies has been assumed to be equivalent to that of the bovine enzyme. were trypsinized from the flasks and washed twice in PBS. All subse- Extensive work to define the gene and protein sequences of the sub- quent steps were performed at 4 °C. The cells were resuspended in units of the human enzyme proceeded by identifying homologues to ~4.5 mL of buffer containing 20 mM HEPES (pH 7.8), 5 mM KCl, −1 the bovine sequences [16]. Then, in 2003, Murray and coworkers used 1.5 mM MgCl2,1mgmL BSA and a protease inhibitor cocktail immuno-purified human complex I to detect 42 homologues to (Roche), incubated for 10 min, then disrupted by seven passes through known bovine proteins [17] and, in 2005, Schilling and coworkers a cell homogenizer (Isobiotec) fitted with a 12 μm ball. For every 3 mL of isolated complex I from mouse brain tissue and from cultured cells lysed cells, 2 mL of buffer containing 20 mM HEPES (pH 7.8), 525 mM and detected 41 homologues to known bovine proteins [18].Inthe mannitol, 175 mM sucrose, and the protease inhibitor cocktail were same study complex I from a rat cell line was analyzed and 33 homo- added, then the volume made up to 30 mL with MSH buffer (containing logues were identified. Recently, complex I from ovine heart, a close 210 mM mannitol, 70 mM sucrose, 20 mM HEPES (pH 7.8), 2 mM EDTA relative of the bovine complex, has also been shown to contain the and the protease inhibitor cocktail). Cell nuclei were removed by centri- same 44 subunits [19]. However, all of these studies have taken the fugation (750 x g for 10 min), then crude mitochondria were collected bovine heart complex as a model and not searched for species or tissue (8000 x g for 20 min). They were resuspended in MSH, treated for specific subunits or subunit isoforms that may not be present in it. 15 min with 50 U/mL benzonase (Millipore), then layered onto a 1.5– Here, we have characterized the composition of complex I isolated 1–0.5 M sucrose step gradient (also containing 10 mM HEPES (pH 7.8) from five rat tissues: heart, skeletal muscle, kidney, liver and brain. and 5 mM EDTA). Following centrifugation (8500 x g for 60 min), the We used chromatographic protocols developed for the bovine complex brown band was collected, diluted in MSH buffer, and recentrifuged to [20], adapted to small scale for the rat tissues, as well as immuno-puri- collect the mitochondria. fication from the same set of tissues and also cultured cells. The aims were to determine whether all 44 different subunits of the bovine en- 2.4. Purification of complex I by chromatography zyme are present in each rat tissue, and to search for additional proteins or isoforms that are associated with the complex, perhaps in a tissue- The following method was adapted from that of Sharpley et al. [20]. dependent fashion. We report discovery of a new isoform for subunit Mitochondrial membranes (~5 mg mL−1) were solubilized by addition NDUFV3 that, with a molecular mass of ~50 kDa, is more than five of 0.8–1.2% lauryl maltose neopentyl glycol (LMNG, Anatrace) for times longer than the canonical ~10 kDa form. 30 min. Insoluble materials were removed by centrifugation (4800 x g for 30 min) then the supernatant was filtered (0.22 μm polyethersul- 2. Materials and methods fone syringe filter) and loaded onto a Q-sepharose HP column (GE Healthcare) pre-equilibrated with buffer A (20 mM Tris-HCl (pH 7.7), 2.1. Cell culture 10% ethylene glycol, 0.1% LMNG, 1 mM EDTA, 0.005% soy bean asolectin (Avanti Polar Lipids) and 0.005% CHAPS (Santa Cruz Biotechnology). The MH-TC-5123 line was purchased from CLS Cell Lines Service The column was washed with 22 or 24% buffer B (buffer A plus 1 M GmbH, and the U2OS line (from Professor M. Ashcroft, Cambridge) NaCl) then complex I was eluted at 33 or 35% buffer B using either a lin- was authenticated by Eurofins Genomics; all cells were confirmed neg- ear or step elution. To remove F1Fo-ATPase, complex-I containing frac- ative for mycoplasma. MH-TC-5123 cells were cultured in RPMI medi- tions were incubated for 15 min with 1 mM ATP, 1 mM MgSO4 and um (Gibco) supplemented with 4.5 g L−1 glucose, 10% FBS (Hyclone) 0.25 μg of a modified version of the bovine ATPase inhibitor protein and 1 mM glutamine. U2OS cells were cultured in DMEM medium IF1, carrying a C-terminal histidine tag [25]. The samples were passed (Gibco), supplemented similarly. through a HisTrap HP column (GE Healthcare), pre-equilibrated with buffer containing 20 mM Tris-HCl (pH 7.7), 10% ethylene glycol, 0.1% 2.2. Isolation of mitochondria and mitochondrial membranes from rat LMNG, 1 mM EDTA and 20 mM NaCl, collected and concentrated. Final- tissues ly, the sample was applied to a 2.4 mL Superose 6 size exclusion column run in buffer containing 20 mM Tris-HCl (pH 7.7), 10% ethylene glycol, Mitochondria from the tissues of albino Wistar rats were prepared 0.02% LMNG, and 150 mM NaCl; the complex I-containing fractions using protocols adapted from Chapel and Hansford [21] and Tyler and (which eluted at ~1.3 mL) were collected and concentrated for storage Gonze [22]. All the steps following tissue excision were at 4 °C. Briefly, at −80 °C. the tissues were diced then homogenized. Skeletal muscle and heart were homogenized in buffer containing 30 mM Tris-HCl (pH 7.4), 2.5. Purification of complex I by immuno-purification 225 mM mannitol, 75 mM sucrose and 0.5 mM EGTA by electrical homogenization, and brain, liver and kidney in 5 mM Tris-HCl (pH 7.4), Complex I was immuno-purified from rat tissue mitochondria using 250 mM sucrose and 2 mM EDTA in a Dounce homogenizer. Debris a commercial kit (Abcam ab-109,711) according to the manufacturers was removed by centrifugation (1500 x g for 5 min) and by using a mus- instructions, and from cell line MH-TC-5123 by the same method but lin cloth. Then, crude mitochondria were collected by centrifugation following the digitonin treatment of Andrews et al. [26]. All steps (10,000 x g for 25 min), resuspended in buffer containing 5 mM were carried out at 4 °C. Briefly, mitochondria and digitonin-treated HEPES (pH 7.4), 250 mM mannitol and 0.5 mM EGTA, and layered cells were solubilized with 1% n-dodecyl β-D-maltoside (DDM), centri- onto 30% Percoll medium containing 25 mM HEPES (pH 7.4), 225 mM fuged to remove insoluble material, and incubated with the affinity mannitol and 1 mM EGTA. The samples were centrifuged (95,000 x g beads overnight. The beads were washed twice before complex I was for 30 min) and the mitochondria (in the lower brown band) collected eluted in buffer containing 200 mM glycine (pH 2.5) and 0.05% DDM. and resuspended in buffer containing 20 mM Tris-HCl (pH 7.7), 10% The pH was neutralized by addition of Tris base. H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 199

2.6. Electrophoresis peptide precursor mass tolerance of 5 ppm and fragment mass tolerance of 0.01 Da (0.5 Da for data from the Orbitrap XL), allowing for up Samples were reduced with dithiothreitol or tris(2- to four missed cleavages and variable modifications (methionine oxida- carboxyethyl)phosphine and analyzed by SDS PAGE using tris-glycine tion, protein N-formylation and N-acetylation). For analysis of peptides 10–20% acrylamide gels (Life Technologies) with bovine complex I extracted from SDS PAGE gel slices by matrix-assisted laser-desorption and a molecular weight ladder (Kaleidoscope, Bio-Rad) for comparison. ionization (MALDI), an Applied Biosystems/MDS SCIEX model 4800 Plus Mitochondria from U2OS cells were analyzed by Blue Native PAGE (3– MALDI–TOF-TOF spectrometer was used [29]. Spectra were assigned 12% acrylamide gels, Life Technologies) at 4 °C according to the manu- using the same variable modifications plus cysteine propionamide and facturers instructions; after 1 h the standard cathode buffer was allowing for up to three missed cleavages, with a peptide mass tolerance replaced with one containing 10% of the Coomassie concentration. All of 360 ppm and a fragment mass tolerance of 0.8 Da. Only peptide as- gels were stained with Coomassie R-250. signments made by Mascot scoring at or above its P b 0.05 threshold were considered. 2.7. Creation of a custom database of rat complex I subunit sequences 2.9. Blue native PAGE profiling, with automated relative peptide intensity A custom sequence database, based on the Ensembl (http://www. analyses ensembl.org) database Rnor_6.0.all (July 2015), was created in order to search mass spectra generated from rat proteins. All the sequences Profile analyses of mitochondria from U2OS cells were carried out for complex I subunits were annotated with meaningful descriptions, following the method of Heide et al. [30]. Equally-sized slices from a and aligned with their well-characterized bovine counterparts to BN-PAGE lane were digested with trypsin, peptides extracted by addi- inspect their presence, their sequence homologies and their lengths. tion of 60% acetonitrile, 4% formic acid, and a portion of each extraction Sequences for rat subunits ND2 and NDUFS6 were absent from the dried down to completion and resuspended in 2% acetonitrile, 0.1% downloaded resource so, respectively, NCBI sequences (https://www. formic acid. The peptides in each slice were analyzed sequentially ncbi.nlm.nih.gov/protein) AP_004893.1 and NP_062096.1 were added. using the Q-Exactive Orbitrap Plus instrument as described above. For The sequence for rat subunit NDUFB1 was also absent, but the only peptide sequence assignment, ProteomeDiscoverer (Thermo Fisher Sci- candidate replacement protein sequences in the NCBI database were entific) software was used to submit spectra to Mascot 2.4 (Matrix Sci- derived from low quality DNA sequences and hence contained undeter- ence Ltd.) configured to use a Uniprot (http://www.uniprot.org) Homo mined residues. Therefore, the well-characterized bovine sequence was sapiens database (UP00005640, December 2015), augmented with the used to search the rat chromosomal sequences with the tblastn applica- sequence of the long isoform of NDUFV3 (termed NDUFV3L). Search pa- tion (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=tblastn). Two rameters included the same variable peptide modifications as above, short regions of homologous protein were identified in different reading with cysteine propionamide and allowing for the presence of only one frames on chromosome 6. The chromosomal sequence covering both missed cleavage, and a peptide mass tolerance of 10 ppm and fragment regions was examined using the Augustus gene prediction program ion mass tolerance of 0.5 Da. ProteomeDiscoverer counted the abun- [27], and the predicted protein sequence (found to be identical to EST dances for all the well-assigned peptides observed in all the samples, sequence AI170469.1 from a R. norvegicus lung sample) was added to and its standard aggregation algorithm was employed as a measure of the database. relative protein abundance for each protein for each gel slice; up to The custom database was further augmented with the predicted the three highest intensity peptides for each protein were selected mRNA splice variants of complex I subunit-encoding genes. Sequences and averaged. were retrieved from the NCBI Aceview mRNA resource [28] and the Au- gustus application [27] was used to predict further hypothetical isoform 2.10. Semi-automated relative peptide intensity analyses variants from portions of the chromosomal DNA sequences for each gene from Rnor_6.0, extended with 10 kbp of the flanking regions. The The raw mass spectra from LC-MS analyses were analyzed in greater procedure was carried out using the parameters for predicting human detail specifically for evidence pertaining to the long and short isoforms splice variants. In total, 182 alternative isoforms for 34 subunits were of NDUFV3. Xcalibur Qual Browser (version 3.0.63, Thermo Fisher added to the database; no alternative transcripts were predicted for Scientific) was used to create extracted ion chromatograms (XICs) for subunits NDUFA2, NDUFA13 or NDUFB10. The custom database was peptides identified previously with Mascot in various ion charge states. uploaded to an in-house Mascot 2.4 server (Matrix Science Ltd.) XICs were created and inspected for the 2+,3+ or 4+ charge states of various ions from complex I subunit peptides, with mass tolerances of 2.8. Mass spectrometry analyses 20 ppm. The peak volume observed at the expected retention time of the peptide was calculated using Xcalibur's integration tool, and the Protein samples for mass spectrometry analyses were either fragmentation spectrum of the peptide was examined to validate its precipitated from solution with cold ethanol and digested in 50 mM am- identity. The sums of the intensities for peptides unique to the long monium bicarbonate with trypsin, or digested in-gel in 20 mM Tris-HCl and short isoforms were normalized to the sum of the intensities for

(pH 8) and 5 mM CaCl2, with subsequent peptide extraction. For LC-MS the common peptides, to allow comparison between samples. analyses, samples were fractionated on an Acclaim PepMap nanoViper C18 reverse-phase column (Thermo Scientific) (75 μm × 150 mm), 2.11. Bioinformatic analyses with a gradient of 5–40% acetonitrile in 0.1% formic acid at a flow rate of 300 nL min−1 over 84 min, then peptides were analyzed by a Q- Homologues to the human sequences for the long and short isoforms Exactive Plus Orbitrap mass spectrometer (Thermo Scientific) with frag- of NDUFV3 were identified by tBlastn searches against the NCBI RNA mentation performed by higher-energy collisional dissociation (HCD) RefSeq database using the BLOSUM62 scoring matrix. Sequences that using nitrogen. The mass range was 400 to 1600 m/z for the precursor had an expect value of b10−4 were considered to be homologous. Sec- ions; the top 10 most abundant ions were selected for MS/MS analyses. ondary structure analyses were carried out using PSIpred [31] and One sample, the in-solution tryptic digest of a chromatographically pu- PredictProtein [32], and GlobPlot 2.3 [33], Disprot [34], and DisEMBL rified complex I from liver, was analyzed with an LTQ Orbitrap XL mass [35] were used to search for regions of intrinsic protein disorder. Se- spectrometer by the closely-related method described by Andrews et al. quence motifs were investigated using NCBI-CDD [36] and Pfam [37]. [26]. Spectra were assigned to peptide sequences and originating pro- N-terminal cleavage sites for mitochondrial import sequences were pre- teins using the Mascot 2.4 application (Matrix Science Ltd.) with a dicted by using Mitoprot [38], TargetP [39], and by comparison to the 200 H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 known bovine cleavage sites. Canonical R-2, R-3 or R-10 motifs for mi- samples, particularly the brain and kidney samples. Complex I was tochondrial processing peptidase (MPP) cleavage [40] were detected also isolated from the same tissues by immuno-purification, using a for all subunits with cleaved import sequences except for NDUFS7, commercial kit (Fig. 1B). Again, the banding patterns for all five samples NDUFS8, NDUFA10 and NDUFV3. are similar, but clear extra bands are apparent in the immuno-purified samples; the strong band at ~25 kDa originates from the antibodies 3. Results used in the affinity kit, and others are due to non-complex I proteins that bind adventitiously to it. For comparison with the enzyme from 3.1. Purification of complex I from five rat tissues tissues, the complex was also isolated from the rat hepatoma cell line MH-TC-5123 by immuno-purification. Complex I was isolated from mitochondrial membranes prepared from five different rat tissues (heart, skeletal muscle, liver, kidney 3.2. Identification of known mammalian complex I subunits in the rat and brain) by adapting to a smaller scale chromatographic methods complexes established for complex I from bovine heart mitochondria [20].Ingen- eral, the rat enzymes were less stable than the bovine enzyme, being Each of the rat complexes were analyzed by mass spectrometry to more prone to aggregation upon solubilization, and the resulting prep- detect the presence of homologues to the 44 different subunits present arations were less pure; modifications made to address these problems in the highly-characterized enzyme from bovine heart [5,8,11,41]. First, included the use of a different detergent, lauryl maltose neopentyl gly- samples of chromatographically- and immuno-purified complex I from col (LMNG) for soubilization and chromatography, and inclusion of an all five tissues were digested in solution with trypsin, and the peptides affinity step to remove ATP synthase. Fig. 1A shows a typical SDS were analyzed by LC-MS. The results are summarized in Fig. 2A, and PAGE analysis of the complex I samples isolated from each tissue, and presented in full in Supplementary Tables 1 and 2. In addition, the com- compares them to a sample of the bovine heart enzyme. It is clear plete mass spectrometry data are given in Supplementary Dataset 1. For from the banding pattern that the subunit composition is similar in all the chromatographically-purified complexes, homologues of 42 bovine cases, although the positions of some of the bands vary between the subunits were detected in the samples from heart and skeletal muscle, species, and a higher level of contamination is present in the rat 41 in the samples from liver and brain, and 43 in the sample from kidney. For the immuno-purified complexes, 41 subunits were detected BSL Bt H MK Markers in the sample from heart, 42 in the samples from skeletal muscle, A brain and kidney, and 40 in the sample from liver. The detection of 75 kDa any subunit relied on at least one peptide scoring above the Mascot 95% confidence threshold. Fig. 2A uses a heatmap to compare the num- NDUFV3L 50 kDa ber of peptides observed for each subunit with the number of tryptic peptide ions with 2+ and 3+ charges predicted by the PeptideMass ap- 37 kDa plication [42] to be possible within the mass range (400–1600 m/z), and 25 kDa allowing for up to four missed cleavages. Although the seven highly-hydrophobic mitochondrial-encoded subunits were not all detected in all 20 kDa samples, they are core components of the enzyme complex and their presence in the enzyme is not in doubt. The presence of every one of the 30 known supernumerary subunits in each of the five tissues was established, with only two missed identifications (NDUFC1 in the chro- 15 kDa matographically-purified liver enzyme, and NDUFA11 in the chromatographically-purified kidney enzyme) of small subunits with relatively few tryptic peptides. The immuno-purified complex from the MH-TC- 5123 cancer cell line was analyzed similarly and 34 subunits were de- 10 kDa tected. In addition to six of the hydrophobic ND subunits, subunits NDUFA1, NDUFA11, NDUFAB1 and NDUFC1 were not detected. LBCL K HSM Bt Markers Peptides characterized from the rat complex I samples were suffi- B 75 kDa cient to define the N-terminal modifications and the mitochondrial target peptide cleavage sites for 31 subunits. Table 1 summarizes the N- NDUFV3L 50 kDa terminal modifications observed here (the complete data are given in Supplementary Table 3), alongside those determined previously in the 37 kDa highly-characterized bovine enzyme [11,43]. Only one minimal differ- ence was identified: for rat NDUFB1, proteins both with and without the initiator methionine were detected. On this basis it is likely that 25 kDa the N-terminal processing of the remaining 13 rat subunits also match their bovine homologues: the predicted modifications and cleavage 20 kDa sites are included in Table 1.

15 kDa 3.3. Additional proteins detected in the rat complex I samples

10 kDa In order to investigate the possibility that extra species- or tissue- specific subunits are present in one or more of the rat complexes when compared with the highly-characterized complex from bovine heart, the additional proteins detected by LC-MS mass spectrometry Fig. 1. SDS PAGE analyses of isolated rat complex I preparations. A) Samples purified by were considered. Many of them, such as subunit-α of F1-ATP synthase, chromatography. B) Samples purified by immuno-purification. The boxes show the gel have well-characterized alternative known functions and are common regions where NDUFV3L was detected by MALDI-TOF analyses. B, brain; L, liver; H, heart; SM, skeletal muscle; K, kidney; Bt, complex I isolated from bovine heart; CL, contaminants of complex I preparations; we do not assign a biological sample from the rat MH-TC-5123 cell line. significance to their presence. However, a considerable number of H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 201 AB

Fig. 2. Heatmap summary of the identification of complex I subunits. The number of theoretical tryptic peptides for each subunit (within the measured mass range) are presented alongside the number detected in each sample. A) summarizes LC-MS data from unfractionated digests and B) summarizes MALDI-TOF data obtained following SDS PAGE fractionation. C, chromatographically purified sample; IP, immuno-purified sample; CL, sample from the rat MH-TC-5123 cell line.

other proteins were also detected (Supplementary Dataset 1). Here, we immuno-purified samples, ACAD9, ACADVL, ECSIT and TMEM126A are focus only on those proteins (see lower panels of Fig. 2)thathavebeen the most likely candidates to interact with the mature complex. associated with complex I previously. Table 2 shows that only ACAD9, To investigate these proteins more thoroughly two additional ap- ECSIT and TMEM126A, proteins involved in the assembly of complex I proaches were used. First, the LC-MS analyses described are highly sen- [44], were detected in both chromatographically- and immuno-purified sitive and detect many proteins present at extremely low levels: if a samples. AIF1 (which has recently been linked to the MIA40 pathway protein can also be identified by MALDI-TOF mass spectrometry, a rela- for oxidative folding of the three CHCH domain-containing subunits of tively less sensitive technique with lower dynamic range, then it is likely complex I [45,46]) and FOXRED1, NDUFAF3 and NDUFAF4 (which are present in a more substantial amount. The chromatographically-puri- also complex I assembly factors [44]) were detected in only chromato- fied samples (containing the most candidate proteins) were resolved graphically-purified samples, along with mitochondrial lactamase-β by SDS PAGE, each gel lane was cut into 50–60 slices, then the proteins (lactB), which has been linked to complex I by phylogenetic profiling present in each section were analyzed by MALDI-TOF (the compositions [47]. NDUFAF1 (a further complex I assembly factor [44]), Isu1 and of unfractionated digests of the complex are too complicated for mean- Nfs1 (involved in the assembly of iron-sulfur clusters [48])andsix ingful MALDI-TOF analysis). Fig. 2B summarizes the MALDI-TOF data, LYR proteins were detected only in immuno-purified samples; the LYR for comparison with the LC-MS data in Fig. 2A. A substantial number proteins share an amino-acid motif with subunits NDUFB9 and of the known subunits were detected (37 in the sample from heart, 34 NDUFA6 [49,50] so it is likely that they interact directly with the immu- in the sample from skeletal muscle, and 35 in the samples from liver, noaffinity matrix and are artificially enriched. As contaminant proteins kidney and brain, all detected in gel sections consistent with their are less likely to be present in both chromatographically- and masses) but the only complex I-associated protein detected was lactB 202 H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207

Table 1 Table 2 N-terminal modifications and mitochondrial target peptide cleavage sites for the sub- Summary of mass spectrometry evidence for the long and short NDUFV3 isoforms. Rat units of mammalian complex I. Information for B. taurus was taken from [11,43]. Infor- samples purified by chromatography (C) or immuno-purification (IP) were analyzed by mation from rat is experimentally determined unless in square brackets, in which case it LC-MS and MALDI-TOF mass spectrometry. Peptides from the human cell line U2OS are has been predicting by combining data from B. taurus with sequence-based prediction of those found in the gel slice containing intact complex I in the blue native PAGE analysis. the cleavage sites. Mascot scores for each identification are presented, compared to the p b 0.05 ions score cut-off for each experiment. Subunit B. taurus R. norvegicus Number peptides ND1 Formyl Formyl observed Score ND2 Formyl [Formyl] ND3 Formyl Formyl Sample Long Short Common Long Short ND4 Formyl Formyl Heart LC-MS (IP) 2 3 5 731/22 720/22 ND5 Formyl Formyl LC-MS (C) 0 1 2 1761/22 2034/22 ND6 Formyl [Formyl] MALDI (IP) 0 0 0 0 0 ND4L Formyl Formyl MALDI (C) 0 0 0 0 0 NDUFA1 no mod [no mod] Skeletal muscle LC-MS (IP) 2 3 5 1397/22 1554/22 NDUFA10 Δ1–23 Δ1–35 LC-MS (C) 0 2 2 1438/22 1620/22 NDUFA11 -Met + Acetyl [− Met + Acetyl] MALDI (IP) 0 0 0 0 0 NDUFA12 +Acetyl + Acetyl MALDI (C) 0 0 0 0 0 NDUFA13 -Met + Acetyl -Met + Acetyl Liver LC-MS (IP) 5 2 2 580/22 423/22 NDUFA2 -Met + Acetyl -Met +Acetyl LC-MS (C) 3 1 2 220/22 222/22 NDUFA3 -Met + Acetyl [−Met + Acetyl] MALDI (IP) 0 0 0 0 0 NDUFA5 -Met + Acetyl [−Met + Acetyl] MALDI (C) 0 0 0 0 0 NDUFA6 -Met + Acetyl -Met + Acetyl Kidney LC-MS (IP) 1 3 3 781/21 978/21 NDUFA7 -Met + Acetyl -Met + Acetyl LC-MS (C) 0 0 1 33/22 33/22 NDUFA8 -Met -Met MALDI (IP) 0 0 0 0 0 NDUFA9 Δ1–35 [Δ1–35] MALDI (C) 0 0 0 0 0 NDUFAB1 Δ1–68 Δ1–68 Brain LC-MS (IP) 18 2 2 1211/22 318/22 NDUFB1 -Met -Met (partial) LC-MS (C) 5 2 2 702/22 457/22 NDUFB10 -Met -Met MALDI (IP) 0 0 0 0 0 NDUFB11 Δ1–29 Δ1–29 MALDI (C) 1 0 0 42/34 0 NDUFB2 Δ1–36 Δ1–33 MH-TC-5123 LC-MS (IP) 39 4 0 2448/23 109/23 NDUFB3 -Met + Acetyl (partial) -Met + Acetyl (partial) cell line MALDI (IP) 3 0 0 228/31 0 NDUFB4 -Met + Acetyl -Met + Acetyl U2OS cell line LC-MS 10 1 0 748/25 109/25 NDUFB5 Δ1–46 [Δ1–46] (complexome) NDUFB6 -Met + Acetyl -Met + Acetyl NDUFB7 -Met + Myristoyl -Met + Myristoyl NDUFB8 Δ1–28 [Δ1–28] NDUFB9 -Met + Acetyl -Met + Acetyl NDUFC1 Δ1–27 [Δ1–27] or more peptides also present in the canonical isoform were discarded. NDUFC2 +Acetyl + Acetyl Only one splice variant was unambiguously identified: a ‘long’ isoform Δ – Δ – NDUFS1 1 23 1 23 of subunit NDUFV3 (NDUFV3L). The canonical ‘short’ NDUFV3 subunit NDUFS2 Δ1–33 [Δ1–33] NDUFS3 Δ1–38 [Δ1–36] (NDUFV3S) has a mass of 8.2 kDa and is generally referred to as the NDUFS4 Δ1–42 Δ1–42 10 kDa subunit. The long isoform has a mass of 45.6 kDa and so, by anal- NDUFS5 -Met -Met ogy, we refer to it as the 50 kDa isoform. Fig. 3 compares the gene struc- NDUFS6 Δ1–28 [Δ1–20] ture and protein sequences for the two isoforms; exons 1 and 4 are Δ – Δ – NDUFS7 1 37 1 37 common to both, but exon 3 is only expressed in the long version. No NDUFS8 Δ1–36 Δ1–34 NDUFV1 Δ1–20 Δ1–20 evidence for an alternative isoform of NDUFB11, as detected previously NDUFV2 Δ1–32 Δ1–31 in cultured human cells [51],wasidentified. NDUFV3 Δ1–34 Δ1–35 Peptide evidence obtained for both NDUFV3 isoforms is summarized in Table 2. In LC-MS analyses of the chromatographically-purified samples, peptides unique to the long form of NDUFV3 were observed only in inthesamplepurified from liver. Second, we interrogated existing blue- complex I from liver and brain, whereas peptides unique to the short native PAGE profiling data on mitochondria from the human osteosar- isoform, or common to both, were observed in the samples from all tis- coma cell line U2OS, to search for proteins that co-migrate with com- sues. In LC-MS analyses of the immuno-purified samples, peptides plex I. None of the proteins investigated showed a distinct intensity unique to both the short and long isoforms, as well as those common peak co-incident with mature complex I. In summary, it is highly unlike- to both, were observed in all samples. Thus, the long isoform was de- ly that any of the additional proteins detected are specifically associated tected in complex I samples purified by both methods. Considerably with mature complex I in any of the cells or tissues studied, suggesting more peptides derived from the long isoform were detected in the sam- they are simply low-level contaminants or they are present in partially- ple isolated from cultured cells. Although the number of peptides de- assembled intermediates. tected is not a reliable quantitative measure of protein abundance, the data suggest that the long isoform is present at higher abundances in 3.4. Identification of an alternative isoform of subunit NDUFV3 in complex I brain and liver, and that it is significantly upregulated in cultured cells. (NDUFV3L) In MALDI-TOF analyses, which as mentioned are much less sensitive, the long isoform was detected only in chromatographically-purified To search for alternative splice variants of the complex I subunits, complex I from brain, and numerous peptides were detected in the sam- two isoform prediction programs were used to construct the protein se- ple from cultured cells (see Table 2). Crucially, in both cases the peptides quences of hypothetical alternative isoforms (a total of 182 isoforms originated from the ~55–60 kDa region of the SDS PAGE analysis (see were predicted for 34 subunits). These sequences were added to the da- Fig. 1) where a band is visible in the lane containing the sample from tabase of rat protein sequences, and compared with the LC-MS dataset. cultured cells. In support of a higher level of the long isoform in brain Discerning between isoforms present on the basis of peptide data is than in the other tissues, LC-MS analyses of the same ~55–60 kDa region complicated by the fact that many peptides are common to multiple iso- from SDS PAGE analyses of mitochondria isolated from each tissue de- forms. Thus, identifications of isoforms made solely on the basis of one tected the long isoform only in the sample from brain. H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 203

Finally, Fig. 4 shows the relative intensity profiles for NDUFV1, in the ovine complex, so it is likely that these two subunits are present NDUFV3 and NDUFV3L in a blue native PAGE analysis of mitochondria in the porcine complex, but that their density is only poorly resolved. from human U2OS cells. The peak for NDUFV1 (a core subunit of com- Notably, the subunit compositions of the B. taurus, S. scrofa and O. aries plex I) in slice 19 corresponds to the position of mature complex I, complexes were all determined using intact and functional chromato- and therefore it is clear that both the long and short isoforms of graphically-purified material. Previously, mass spectrometry analyses NDUFV3 are associated with the intact complex. A proportion of of the human and rodent complexes [17,18] have been performed on NDUFV3L is also present towards the bottom of the gel, in a position cor- immuno-purified material, and on the rodent enzyme purified using a responding to the monomeric protein. NDUFV3 is located peripherally sucrose gradient, and the majority of the expected subunits were de- in the molecular structure of bovine complex I [14], so it may be only tected. Here, we have identified the complete set of 30 supernumerary weakly bound and dissociate during electrophoresis – or it may also subunits in intact chromatographically-purified rat complex I (a mam- be present as the free protein in mitochondria. In summary, our data malian species less closely related to B. taurus than S. scrofa and O. suggest that NDUFV3L is a bona-fide component of rat complex I, and aries). Together, these analyses confirm the composition of the canoni- present also in the human enzyme, but that it is present in a cal 45-subunit mammalian complex I, supporting different mammalian substoichiometric level that varies between tissues. species being used as models for the human enzyme. Furthermore, identification of the same 44 subunits in five different rat tissues (and 3.5. Estimation of the relative abundance of the long and short isoforms in in cell lines) suggests that the canonical enzyme is not specificto different tissues heart tissue, and identification of the same N-terminal modifications in the rat and bovine enzymes suggests that they are also common The relative abundance of the long and short isoforms of NDUFV3 in throughout mammals. different tissues was evaluated in a low-precision estimation based on unlabeled peptide signal intensities from LC-MS analyses. Two peptides 4.2. The challenge of identifying hitherto-unknown subunits unique to the long isoform, two unique to the short isoform, and two common to both isoforms were selected from the sets of peptides that Identifying the presence of known subunits in a sample of complex I surpassed the 95% Mascot confidence threshold in at least one sample is relatively straightforward, provided that data are searched against a (see Supplementary Table 4). A total of 110 peptide intensities (includ- sufficiently complete and well annotated database. Evaluating whether ing multiply charged ions) were manually assessed in the 11 samples additional subunits are present is a much greater challenge: many addi- investigated, and 100 peaks were quantified. For 73 of them the tional proteins are detected in mass spectrometry analyses but it is dif- fragmentation patterns were matched to the sequence by Mascot, ficult to decide if they are contaminants, proteins that bind transiently confirming the peptide identities. For six of them the fragmentation pat- such as assembly factors, chaperones, and proteins involved in signaling terns were not matched but were similar to those matched by Mascot, or degradation pathways — or subunits. Here, we investigated addition- and the remaining 21 assignments were based on shared peptide al proteins in our preparations by discarding proteins with known alter- masses and retention time only. The peak volumes for the two peptides native functions, comparing data from different methods of isolation unique to the long isoform were summed, as were those for the com- and analyses, and by focusing on proteins with a previous reported mon and short-isoform peptides. To account for variation in sample connection to the complex. We did not identify any new candidate composition, the relative abundance estimate for the long isoform was subunits, in any of the tissues considered — but cannot exclude the pos- obtained by dividing the unique-to-long-isoform summed value by sibility that tissue-specific subunits exist to be identified in the future. that of the common-to-both summed value, and this calculation was made similarly for the short isoform. Then, to provide an estimation of 4.3. The ‘long’ isoform of subunit NDUFV3 in complex I how the relative abundance of each isoform varies between tissues, each value was normalized to the value from immuno-purified heart The long, 50 kDa isoform of NDUFV3 has been found to be associated complex I. The values obtained are given in Fig. 5. Note that they provide with complex I in five different rat tissues, and in both rat and human no information about absolute stoichiometries, only a comparison of the cultured cells, including in complex I samples isolated by three different relative amounts in the different tissues. Consistent with the observa- methods (chromatography, immuno-purification and blue native tions described above, our data suggest that the long isoform is consid- PAGE). The long isoform is most abundant in brain tissue, and particu- erably more abundant, relative to the short isoform, in complex I from larly abundant in cultured cells, and the very low levels detected in cultured cells, and also more abundant (though to a lesser extent) in heart, skeletal muscle, kidney and liver are consistent with the fact complex I from brain tissue. The relative abundance appears to be low that it has never been detected in the extensively characterized enzyme (and similar) in all other tissues; the lack of precision in the estimated from bovine heart. However, even in cultured cells (with the highest values precludes further interpretation. abundance of NDUFV3L) only a faint band is visible in SDS PAGE analyses (Fig. 1), showing that the long isoform is substoichiometric in all 4. Discussion cases, particularly so in heart, skeletal muscle and kidney. Fig. 3 shows that the long, 50 kDa isoform of NDUFV3 shares its N- 4.1. The canonical mammalian complex I and C-termini with the short, 10 kDa (canonical) form, but contains a large insertion between them. Bioinformatic analyses suggest that a 45 subunits were identified and modeled in the structure of complex large portion (predicted 40–90%, depending on the analysis program IfromB. taurus heart mitochondria [13,14,52], substantiating the con- used) of the protein is present as disordered loop structures, no sub- sensus value of 44 different subunits and two copies of the acyl-carrier stantial secondary structural elements are predicted to be present, and protein determined by protein and mass spectrometry analyses [5,8, no strong homology to any other protein can be identified (although 11,41]. The subsequent structure of ovine complex I showed the same motif searches suggested some similarity to ribonuclease E, DNA poly- set of 45 subunits [19], whereas the transmembrane helix from subunit merase II subunits γ and τ, and DNA topoisomerase 2). A notable feature NDUFB3 (as well as subunit NDUFA12 on the hydrophilic arm) were of the structure of the hydrophilic domain of complex I from B. taurus not modeled in the complex I-containing structure of the porcine [14] are the long loops that run across its surface, particularly at subunit respirasome [53]. Previous MALDI-TOF analyses of the complexes I interfaces. A good example is subunit NDUFA7, which arches up from from ovine and porcine hearts (prepared using the method of Sharpley the bottom of the domain along the NDUFS8–NDUFS2 subunit interface, and coworkers) [54] detected all 30 expected supernumerary proteins crosses NDUFS2, runs along its interface with NDUFS1 and then onto in the porcine complex, and only missed identifying subunit NDUFC1 NDUFS3. A long loop from the central portion of NDUFV3 would thus 204 H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 A

Fig. 3. The two isoforms of NDUFV3. A) The gene structure of NDUFV3 in R. norvegicus with the exons shown in red and the introns as black lines. B) Pairwise sequence alignment of the two protein products. The mitochondrial targeting peptide is in grey and two residues that were identified as variants in mitochondrial disease patients [3] and conserved between the human and rat sequences are marked in blue. Residues in bold were detected by peptide mass fingerprinting, and serine residues homologous to phosphorylated residues identified in proteomics studies are marked in orange [57–59]. An unusual acidic region featuring a chain of serine residues in the same region is underlined. be consistent with the structures and forms of other subunits on the 4.4. Conservation of NDUFV3L in other species same domain. In the structure of complex I from B. taurus,NDUFV3islo- cated at the top of the hydrophilic arm, adjacent to NDUFV1 and To investigate the wider prevalence of the NDUFV3 splice variants, NDUFV2 [14], most likely with the conserved C-terminal region in a the NDUFV3L and NDUFV3S sequences were searched against the cleft between NDUFV1 and NDUFV2. It is thus likely that the N-andC- NCBI RNA database (see Fig. 6). NDUFV3 appeared in evolution around termini common to both isoforms anchor the subunit to the complex, the time that lobe-finned fishes developed, as a homologue could not be and that the two isoforms are interchangeable. Stroud and coworkers identified in jawless fish, insects, nematodes or fungi. Analysis of the [55] suggested that in cell lines NDUFV3 is the final subunit to assemble structure of the human NDUFV3 gene revealed the same splicing pat- into the mature complex, and demonstrated that copies of the short iso- tern as in rat (consistent with detection of NDUFV3L in human U2OS form are readily exchanged in and out of the mature complex in vivo. Consequently, we propose that the long and short isoforms compete for a common binding site on complex I (such that each complex contains a single NDUFV3 subunit of either isoform), with their relative A 12 levels of incorporation in different cells and tissues governed by the rel- 10 ative levels of the isoforms present in the mitochondria. 8 6 1 4 2 Relative abundance

0.8 (normalized to heart (IP) 0 HSMK L B

0.6 B 254 252 0.4 250 Relative abundance

0.2 12 10 0 0 10 20 30 40 50 60 8

Gel slice number Relative abundance 6 (normalized to heart (IP) 4 2 0 H SM K L B CL Fig. 4. Blue native PAGE profiling for the NDUFV1, NDUFV3L and NDUFV3S subunits of complex I. The gel slice shown was cut into approximately 60 slices and each slice analyzed by LC-MS. The signal intensities for the different subunits (black for NDUFV1, red for NDUFV3L and green for NDUFV3S) are plotted against the gel, showing that all Fig. 5. Unlabeled peptide quantification of NDUFV3S and NDUFV3L. A) three proteins have their greatest abundance in the same slice, corresponding to the Chromatographically- purified complexes I. B) Immuno-purified complexes I. The short position for intact complex I. Note that the slice width is not precise so the graph scale isoform is in grey and the long isoform in black. H, heart; SM, skeletal muscle; K, kidney; and gel do not match exactly. L, liver; B, brain: CL, rat MH-TC-5123 cell line. H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 205 A

Fig. 6. The NDUFV3 isoforms in different species. A) Phylogenetic tree showing the distribution of the two NDUFV3 isoforms. Species in red boxes do not contain NDUFV3. Species in green boxes do contain NDUFV3 and an asterisk is used to denote those for which the long isoform has been identified. B) Consensus sequences for the two isoforms based on multiple sequences alignments from the species in A. The most common residue at each position is given, and colors indicate the percentage identity. A possible N-terminal extension to the short sequence from Xenopus tropicalis has been omitted as a methionine is also present at the usual position of the start codon; the length of the alignment is extended by insertions in the NDUFV3L sequence from salmon. cells) and the alternative splicing is known to involve the catenin pro- long isoform, are well conserved, but that the rest of the protein se- tein known as ‘Armadillo Repeat gene deleted in Velo-Cardio-Facial syn- quence is only poorly conserved. drome’ (ARVCF) [56]. NDUFV3L itself is present in many species from fish to humans — and it is possible that for some species RNA sequenc- 4.5. The role of NDUFV3L ing has not yet been deep enough to identify it. Sequence alignments for the NDUFV3 homologues identified (see Fig. 6) show that both the C- Very little information pertaining to a physiological role for terminus (conserved in both isoforms) and an unusual region contain- NDUFV3L is available. First, four mutations in NDUFV3 have been ob- ing a string of ten serine residues (see also Fig. 3), particular to the served in patients with mitochondrial disease [3] (see Fig. 3), although 206 H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 they have not been confirmed as causatory. Two of the residues are [12] E. Balsa, et al., NDUFA4 is a subunit of complex IV of the mammalian electron transport chain, Cell Metab. 5 (2012) 378–386. conserved in rat (R26Q and K56N). R26 is part of the mitochondrial [13] K.R. Vinothkumar, J. Zhu, J. Hirst, Architecture of mammalian complex I, Nature 515 import sequence, so the mutation may affect mitochondrial import (2014) 80–84. and N-terminal processing. The codon for K56 is adjacent to the [14] J. Zhu, K.R. Vinothkumar, J. Hirst, Structure of mammalian respiratory complex I, Na- ture 536 (2016) 354–358. exon2-intron2 boundary, however the mutation is not predicted to af- [15] T. Gabaldón, D. Rainey, M.A. Huynen, Tracing the evolution of a large protein com- fect splicing, and its presence at ~0.5% in the ExAC database (http:// plex in the eukaryotes, NADH:ubiquinone oxidoreductase (complex I), J. Mol. Biol. exac.broadinstitute.org), including as a homozygous variant, suggests 348 (2005) 857–870. it is not pathogenic. The two non-conserved residues are in sequence [16] J.L.C.M. Loeffen, et al., cDNA of eight nuclear encoded subunits of NADH:ubiquinone oxidoreductase: human complex I cDNA characterization completed, Biochem. that is specific to the long isoform. Second, phospho-proteome analyses Biophys. Res. Commun. 253 (1998) 415–422. [57–59] have identified three of the conserved serine residues in the [17] J. Murray, et al., The subunit composition of the human NADH dehydrogenase obtained by rapid one step immunopurification, J. Biol. Chem. 278 (2003) central serine-rich motif to be phosphorylated in various conditions – – fi 13619 13622. (see Fig. 3) and indeed NDUFV3L was rst discovered as a 65 kDa mi- [18] B. Schilling, et al., Rapid purification and mass spectrometric characterization of mitochondrial phosphoprotein (MIPP65) in rat tissue, as a substrate for tochondrial NADH dehydrogenase (complex I) from rodent brain and a dopaminer- protein kinase N1 [60]. Third, NDUFV3 is located on chromosome 21 gic neuronal cell line, Mol. Cell. Proteomics 4 (2005) 84–96. [19] K. Fiedorczuk, et al., Atomic structure of the entire mammalian mitochondrial com- and so is upregulated in patients with trisomy 21 Downs syndrome plex I, Nature 538 (2016) 406–410. [61]. Fourth, NDUFV3L has been found to be up-regulated in cortical is- [20] M.S. Sharpley, R.J. Shannon, F. Draghi, J. Hirst, Interactions between phospholipids chemia [62] and this, in conjunction with our finding of a higher level of and NADH:ubiquinone oxidoreductase (complex I) from bovine mitochondria, Bio- – fi chemistry 45 (2006) 241 248. NDUFV3L in brain complex I, may suggest it has a tissue-speci crole.Fi- [21] J.B. Chappell, R.G. Hansford, in: G.D. Birnie (Ed.), Subcellular components: prepara- nally, in its size, lack of secondary structure, and proximity to the flavin tion and fractionation, second ed.Butterworths, London 1972, pp. 77–91. binding site of complex I, NDUFV3L resembles the 68 kDa fragment of [22] D.D. Tyler, J. Gonze, The preparation of heart mitochondria from laboratory animals, Methods Enzymol. 10 (1967) 75–77. an atypical cadherin (Ft4) that has been reported to associate with com- [23] J.E. Walker, J.M. Skehel, S.K. Buchanan, Structural analysis of NADH:ubiquinone ox- plex I in Drosophila [63]. No evidence has been found for any interaction idoreductase from bovine heart mitochondria, Methods Enzymol. 260 (1995) between any cadherin protein and mammalian complex I and, although 14–34. a Drosophila homologue to NDUFV3 has been proposed previously [15], [24] M. Minczuk, et al., TEFM (c17orf42) is necessary for transcription of human mtDNA, Nucleic Acids Res. 39 (2011) 4284–4299. we have not been able to identify any high-scoring match sequence. [25] M.J. Runswick, et al., The affinity purification and characterization of ATP synthase Therefore, it is possible that NDUFV3L and Ft4 play similar roles in com- complexes from mitochondria, Open Biol. 3 (2013) 120160. plex I. [26] B. Andrews, J. Carroll, S. Ding, I.M. Fearnley, J.E. Walker, Assembly factors for the membrane arm of human complex I, Proc. Natl. Acad. Sci. U. S. A. 110 (2013) Supplementary data to this article can be found online at http://dx. 18934–18939. doi.org/10.1016/j.bbabio.2016.12.001. [27] M. Stanke, et al., AUGUSTUS: ab initio prediction of alternative transcripts, Nucleic Acids Res. 34 (2006) W435–W439. [28] D. Thierry-Mieg, J. Thierry-Mieg, AceView: a comprehensive cDNA-supported gene Transparency document and transcripts annotation, Genome Biol. 7 (Suppl. 1) (2006) S12. [29] H.R. Bridges, I.M. Fearnley, J. Hirst, The subunit composition of mitochondrial NADH: The Transparency document associated with this article can be ubiquinone oxidoreductase (complex I) from Pichia pastoris, Mol. Cell. Proteomics 9 (2010) 2318–2326. found, in the online version. [30] H. Heide, et al., Complexome profiling identifies TMEM126B as a component of the mitochondrial complex I assembly complex, Cell Metab. 16 (2012) 538–549. [31] D.W. Buchan, et al., Scalable web services for the PSIPRED Protein Analysis Work- Acknowledgements bench, Nucleic Acids Res. 41 (2013) W349–W357. [32] G. Yachdav, et al., PredictProtein - an open resource for online prediction of protein – We thank John E. Walker (MBU) for supplying the IF1 plasmid, Tracy structural and functional features, Nucleic Acids Res. 42 (2014) W337 W343. Prime (MBU) for supplying rat tissues, Shujing Ding (MBU) for [33] R. Linding, R.B. Russell, V. Neduva, T.J. Gibson, GlobPlot: Exploring protein sequences for globularity and disorder, Nucleic Acids Res. 31 (2003) 3701–3708. performing LC-MS analyses and Ian M. Fearnley (MBU) for mass spec- [34] Z. Obradovic, et al., Predicting intrinsic disorder from amino acid sequence, Proteins trometry advice. This work was supported by The Medical Research 53 (2003) 566–572. Council (grant number U105663141 to JH). [35] R. Linding, et al., Protein disorder prediction: implications for structural proteomics, Structure 11 (2003) 1453–1459. [36] A. Marchler-Bauer, et al., CDD: conserved domains and protein three-dimensional References structure, Nucleic Acids Res. 41 (2013) D348–D352. [37] R.D. Finn, et al., Pfam: the protein families database, Nucleic Acids Res. 42 (2014) [1] J. Hirst, Mitochondrial complex I, Annu. Rev. Biochem. 82 (2013) 551–575. D222–D230. [2] E. Fassone, S. Rahman, Complex I deficiency: clinical features, biochemistry and mo- [38] M.G. Claros, P. Vincens, Computational method to predict mitochondrially imported lecular genetics, J. Med. Genet. 49 (2012) 578–590. proteins and their targeting sequences, Eur. J. Biochem. 241 (1996) 779–786. [3] S.E. Calvo, et al., High-throughput, pooled sequencing identifies mutations in NUBPL [39] O. Emanuelsson, S. Brunak, G. von Heijne, H. Nielsen, Locating proteins in the cell and FOXRED1 in human complex I deficiency, Nat. Genet. 42 (2010) 851–858. using TargetP, SignalP and related tools, Nat. Protoc. 2 (2007) 953–971. [4] J.E. Walker, The NADH:ubiquinone oxidoreductase (complex I) of respiratory [40] O. Gakh, P. Cavadini, G. Isaya, Mitochondrial processing peptidases, Biochim. chains, Q. Rev. Biophys. 25 (1992) 253–324. Biophys. Acta 1592 (2002) 63–77. [5] J.E. Walker, et al., Sequences of 20 subunits of NADH:ubiquinone oxidoreductase [41] J. Carroll, I.M. Fearnley, R.J. Shannon, J. Hirst, J.E. Walker, Analysis of the subunit from bovine heart mitochondria. Application of a novel strategy for sequencing pro- composition of complex I from bovine heart mitochondria, Mol. Cell. Proteomics 2 teins using the polymerase chain reaction, J. Mol. Biol. 226 (1992) 1051–1072. (2003) 117–126. [6] J.M. Skehel, I.M. Fearnley, J.E. Walker, NADH:ubiquinone oxidoreductase from bo- [42] M.R. Wilkins, et al., Detailed peptide characterisation using PEPTIDEMASS - a world- vine heart mitochondria: sequence of a novel 17.2-kDa subunit, FEBS Lett. 438 wide-web-accessible tool, Electrophoresis 18 (1997) 403–408. (1998) 301–305. [43] J. Carroll, et al., The post-translational modifications of the nuclear encoded subunits [7] A. Chomyn, M.W.J. Cleeter, C.I. Ragan, M. Riley, R.F. Doolittle, G. Attardi, URF6, last of complex I from bovine heart mitochondria, Mol. Cell. Proteomics 4 (2005) unidentified reading frame of human mtDNA, codes for an NADH dehydrogenase 693–699. subunit, Science 234 (1986) 614–618. [44] S. Guerrero-Castillo, et al., The assembly pathway of mitochondrial respiratory chain [8] J. Carroll, et al., Bovine complex I is a complex of forty-five different subunits, J. Biol. complex I, Cell Metab. 25 (2017) 1–12. Chem. 281 (2006) 32724–32727. [45] E. Hangen, et al., Interaction between AIF and CHCHD4 regulates respiratory chain [9] I.M. Fearnley, et al., GRIM-19, a cell death regulatory gene product, is a subunit of biogenesis, Mol. Cell 58 (2015) 1001–1014. bovine mitochondrial NADH:ubiquinone oxidoreductase (complex I), J. Biol. [46] K. Meyer, et al., Loss of apoptosis-inducing factor critically affects MIA40 function, Chem. 276 (2001) 38345–38348. Cell Death Dis. 6 (2015) e1814. [10] J. Carroll, R.J. Shannon, I.M. Fearnley, J.E. Walker, J. Hirst, Definition of the nuclear [47] D.J. Pagliarini, et al., A mitochondrial protein compendium elucidates complex I dis- encoded protein composition of bovine heart mitochondrial complex I: Identifica- ease biology, Cell 134 (2008) 112–123. tion of two new subunits, J. Biol. Chem. 277 (2002) 50311–50317. [48] R. Lill, Function and biogenesis of iron-sulphur proteins, Nature 460 (2009) [11] J. Hirst, J. Carroll, I.M. Fearnley, R.J. Shannon, J.E. Walker, The nuclear encoded sub- 831–838. units of complex I from bovine heart mitochondria, Biochim. Biophys. Acta 1604 [49] H. Angerer, The superfamily of mitochondrial Complex1_LYR motif-containing (2003) 135–150. (LYRM) proteins, Biochem. Soc. Trans. 41 (2013) 1335–1341. H.R. Bridges et al. / Biochimica et Biophysica Acta 1858 (2017) 197–207 207

[50] H. Angerer, Eukaryotic LYR proteins interact with mitochondrial protein complexes, [57] N. Dephoure, et al., A quantitative atlas of mitotic phosphorylation, Proc. Natl. Acad. Biology 4 (2015) 133–150. Sci. U. S. A. 105 (2008) 10762–10767. [51] D. Panelli, et al., The mechanism of alternative splicing of the X-linked NDUFB11 [58] Y. Bian, et al., An enzyme assisted RP-RPLC approach for in-depth analysis of human gene of the respiratory chain complex I, impact of rotenone treatment in neuroblas- liver phosphoproteome, J. Proteome 96 (2014) 253–262. toma cells, Biochim. Biophys. Acta 1829 (2013) 211–218. [59] H. Zhou, et al., Toward a comprehensive characterization of a human cancer cell [52] J. Zhu, et al., Structure of subcomplex Iβ of mammalian respiratory complex I leads phosphoproteome, J. Proteome Res. 12 (2013) 260–271. to new supernumerary subunit assignments, Proc. Natl. Acad. Sci. U. S. A. 112 (2015) [60] M. Kitagawa, H. Mukai, Y. Ono, Molecular cloning and characterization of a novel mi- 12087–12092. tochondrial phosphoprotein, MIPP65, from rat liver, Exp. Cell Res. 235 (1997) 71–78. [53] J. Gu, et al., The architecture of the mammalian respirasome, Nature 537 (2016) [61] R.F. de Coo, P. Buddiger, H.J. Smeets, B.A. van Oost, Molecular cloning and character- 639–643. ization of the human mitochondrial NADH:oxidoreductase 10-kDa gene (NDUFV3), [54] H. R. Bridges, Ph. D. thesis, Cambridge University, (2009). Genomics 45 (1997) 434–437. [55] D.A. Stroud, et al., Accessory subunits are integral for assembly and function of [62] A. Schneider, A. Fischer, C. Krüger, J. Aronowski, Identiﬁcation of regulated genes human mitochondrial complex I, Nature 538 (2016) 123–126. during transient cortical ischemia in mice by restriction-mediated differential dis- [56] U. Rappe, T. Schlechter, M. Aschoff, A. Hotz-Wagenblatt, I. Hofmann, Nuclear ARVCF play (RMDD), Brain Res. Mol. Brain Res. 124 (2004) 20–28. protein binds splicing factors and contributes to the regulation of alternative splic- [63] A. Sing, et al., The atypical cadherin Fat directly regulates mitochondrial function ing, J. Biol. Chem. 289 (2014) 12421–12434. and metabolic state, Cell 158 (2014) 1293–1308.